You are on page 1of 7

Question 1 (a)

The more complex an experiment, the higher the risks of experimental errors. A more
complex experiment is also more costly than a simple one needing more equipment and
replicates and also more time consuming. Simple experiments are known to be a faster
way to obtain results. They can be used as a basis for more complex experiments and
requiring fewer resources. If only one sample exists, simple experiments are the only
type of experiment available. An example of a simple experimental design is the
Completely Randomized Design. It is the simplest of all experimental designs and is
widely used. The basic principles are randomization and replication.

The advantages of the Completely Randomized Design are:

I. The design is very flexible. Any number of treatments and of replicates may be
used.
II. The statistical analysis is very easy even if the numbers of replicates are not the
same for all treatments.
III. The method of analysis remains simple when the results from some units or
from whole treatments are missing or are rejected.
IV. It provides the maximum number of degrees of freedom for the estimation of
the error variance, which increases the sensitivity or precision for experiment
with small number of treatments.

Complex experiments are typically used to test to factors affecting a particular object.
For example assessing the effects of both bedding and fodder quality in milk yield of
cows. This type of experimental design is called factorial where 2 factors are affecting
milk yield is tested. The disadvantages of this complex experimental design are:

As the number of factors increases, the total number of treatments becomes very large.

Large factorials may be difficult to interpret, particularly when interactions are present.
Question 1 (b)

When re doing the experiment, ensure that the new packaging material is being
compared to other packaging material. This will show the effect of different material on
microbial properties and may increase the chance of rejecting the null hypothesis.

Pre-existing packaging may have been on the market for some time already as
compared to the new packaging which has just been made. The old packaging may have
been done a long time ago. Hence, the comparison is not a fair one as one packaging
may be old and the other is new. The packages have to be made on the same day in
order to make it a fair comparison and to get more accurate results.

The comparison must be made over a period of time for example, take the two
packages and test their microbial properties once a week for at least three months. They
both should be tested on the same day. This will make the comparison a fair one.

Both packages have to be stored under the same conditions. That is, both should be
kept in the same freezer under the same freezing temperatures. Temperature of storage
may have an effect on the microbial properties of the minced meat. Since different
freezers have different temperatures, this can affect the results obtained. Hence, it is
better if both packages were stored in the same freezer.

Different types of meat have different types of microbial profiling and hence different
microbial properties. Therefore, when carrying out the experiment, the type of meat
used has to be taken into consideration. It has to be of the same type because if there
are different microbes present in both samples, then there is no comparison that can be
made. For better and accurate comparisons, the minced meat has to be of the same
variety.
Question 2 (a)

Analysis of Variance

Source DF Adj SS Adj MS F-Value P-Value

Markets 2 7.056 3.5279 8.35 0.005

Error 12 5.070 0.4225

Total 14 12.126

Model Summary

S R-sq R-sq(adj) R-sq(pred)

0.650015 58.19% 51.22% 34.67%

Coefficients

Term Coef SE Coef T-Value P-Value VIF

Constant 5.498 0.168 32.76 0.000

Markets

Port-louis 0.423 0.237 1.78 0.100 1.33

Quatre-Bornes 0.544 0.237 2.29 0.041 1.33


Regression Equation

TVC = 5.498 + 0.423 Markets_Port-louis + 0.544 Markets_Quatre-Bornes

- 0.967 Markets_Rose-Hill

Fits and Diagnostics for Unusual Observations

Std

Obs TVC Fit Resid Resid

6 5.708 4.531 1.177 2.03 R

R Large residual

H0 : market has no effect on yield

H1 : market has an effect on yield

F(2,12) = 3.89

ii) Since F= 3.89 s found in the rejection region, the null hypothesis is rejected
and therefore alternate hypothesis is accepted. there is enough evidence that
market has effect on yield. It can be said that microbial profiling vary according
to market.
Comparisons for TVC

Tukey Pairwise Comparisons: Response = TVC, Term = Markets

Grouping Information Using the Tukey Method and 95% Confidence

Markets N Mean Grouping

Quatre-Bornes 5 6.04264 A

Port-louis 5 5.92145 A

Rose-Hill 5 4.53094 B

Means that do not share a letter are significantly different.

Tukey Simultaneous 95% CIs

The grouping data shows that there is significant difference between A and B markets.

Question 2 (b)

i) Non parametric test is a statistical procedure whereby the data does not match a
normal distribution. The data used in a non-parametric test is frequently of ordinal data
type, thus it does not depend on arithmetic properties .Consequently, all tests involving
the ranking of data are non-parametric and also no statement is made.
Table 1.O .Example of Non- parametric test.

Non- parametric test Aim of test

Wilcoxon signed rank test. Comparison of two dependent samples.

Mann-Whitney U-test. Comparison of two independent samples.

Wilcoxon rank sum test.

Friedman test. Comparison of three or more dependent


samples.

Kruskal-Wallis H-test. Comparison of three or more


independent samples.

ii) Kruskal-Wallis Test: TVC versus Markets

Kruskal-Wallis Test on TVC

Markets N Median Ave Rank Z

Port-louis 5 6.057 9.8 1.10

Quatre-Bornes 5 5.889 10.6 1.59

Rose-Hill 5 4.384 3.6 -2.69

Overall 15 8.0
Test

Null hypothesis H₀: All medians are equal

Alternative hypothesis H₁: At least one median is different

H = 7.34 DF = 2 P = 0.025

The null hypothesis is rejected since p-value (0.025) is less than 0.05. There is strong
evidence that medians are statistically significant. So, it shows that not all the group
medians are equal and further testing is required as there might be an effect of the
market on the yield obtained.